Using neural networks, experts from the New Jersey Institute of Technology and their Chinese colleagues were able to teach microbots to adapt to environmental changes by moving through liquid. The researchers reported their success in a journal article. Communication Physics.
Floating microrobots can be used specifically for targeted drug delivery and medical tasks such as microsurgery. However, their maneuverability so far is limited.
With the help of neural networks and reinforcement learning, the researchers were able to ensure that the micro-robots, which consist of three parts connected by pulling joints, can adapt to difficult conditions and move in the right direction in the liquid, regardless of possible interference. The learning process was similar to how a person learns to swim.
When the microrobot acted a certain way, it received feedback on how accurate its actions were. Later, he gradually learned to swim, based on his experience of interacting with his environment.
After such training, the microrobots were able to move along complex trajectories without prior programming. In addition, they managed to reliably navigate in the presence of disturbances in the environment caused by the movements of fluid flows.
The authors concluded that adaptive behavior is critical for the future use of microrobots that swim in complex environments with uncontrollable and unpredictable environmental factors.